Intelligent Algorithms Of Optimizing Transaction Categorization And Join Query For Distributed Database Systems

With the development and application of information technology(IT),the scale of collecting data,storing data,and processing data in various types of information systems has explosively increased,and the requirements for data-servers with higher capabilities of storing,searching and managing data have been increasingly urgent.Distributed database systems(DDBS)have some dominant advantages such as effective concurrency operation,high service availability and easy functions scalability and the like,which have accordingly received more attentions and comprehensive research strives.The well-developed technologies of DDBS have been applied to many large-scale information systems,for which distributed database design and processing query are still basic & key technical problems in a DDBS.In order to improve request-response speed of DDBS,the dissertation focuses on transaction categorization optimization and join query optimization to improve the efficiency of the whole database system and researches on existing technologies.Detailed analysis of transaction categorization strategy based on hierarchical clustering algorithm has been carried out.The dissertation introduces the concept of transaction execution frequency proportion while constructing a transaction categorization model.And then it combines with the idea of ant colony optimization(ACO)algorithm and proposes that next clustering object could be selected by probability,which solves the problem that hierarchical clustering can not be rolled back.After data segment is deployed to site according to the result of transaction categorization optimization,the dissertation takes into account the environment of low bandwidth network and uses the semi-join algorithm to reduce communication cost of data transmission while constructing a query cost model.Compared with other optimization algorithms,genetic algorithm(GA)is applied in the process of join query optimization of DDBS,operations such as genetic coding,population initialization,selection,crossover and mutation are all improved.The simulation experiment demonstrates that improved algorithm reduces both the correlation between transactions in distributed database and time required for the query.Meanwhile,it improves the request-response speed of the system.